We are quickly drifting into a world where “using AI” is becoming synonymous with “subscribing to someone else’s black box.” Most organisations, teams, and individuals are engaging with AI as consumers of finished products rather than as builders or stewards of technology. That feels like a missed opportunity — and, more importantly, a strategic risk. The open source vs proprietary AI debate is often framed as a technical or commercial one. I think it is much more fundamental than that. It is a question of agency: who gets to understand, shape, adapt, and ultimately control the tools that are increasingly mediating our work, decisions, and relationships.
One of the strongest arguments for open source AI is that it invites people to become producers rather than just consumers. Working with open models, open tooling, and transparent infrastructure forces a different kind of literacy. You start to understand how models behave, what their limitations are, how data shapes outputs, and how brittle “intelligence” can be when it meets real-world messiness. That familiarity matters. It creates organisational muscle memory. It builds internal capability. And it shifts the mindset from “What can this tool do for us?” to “What can we build with this, and how should it behave in our context?” In the long run, that difference compounds.
There is also a very pragmatic economic argument. Right now, AI companies are in land-grab mode. Prices are artificially low because the game is about market share, not sustainability. This is a familiar pattern. We saw it with Uber. We saw it with food delivery platforms. We are seeing it again with generative AI APIs. Once the dust settles and the winners consolidate power, prices will rise. Terms will tighten. And switching costs will be high, especially for organisations that have hard-wired hundreds of workflows, agents, and automations into a single proprietary API. At that point, “vendor lock-in” stops being an abstract risk and starts being an operational reality. Open source does not eliminate cost or complexity, but it does keep exit doors open.
Beyond economics and capability, there is a deeper, more political layer to this. I believe in a world where our tech ecosystems are diverse, plural, and benevolent by design, not monocultures optimised for shareholder value. A healthy open source AI landscape creates space for meaningful innovation that is not dictated solely by the incentives of a handful of dominant firms. It supports democratic values in technology: transparency, contestability, and collective stewardship. And it protects individual and organisational agency by making it possible to inspect, adapt, and fork the tools that increasingly shape our cognition and coordination. This is not nostalgia for some idealised hacker past. It is about future-proofing our capacity to choose.
None of this means that proprietary AI has no place. Closed models will often be more polished, more convenient, and sometimes more capable at the frontier. But if we build our entire AI future on rented intelligence, we should not be surprised when the rent goes up, or when the rules change without our consent. The choices we make now about infrastructure, tooling, and literacy will quietly shape the power dynamics of the next decade. Open source is not just a technical preference. It is a long-term governance strategy.
Some interesting reading and watching
Top 10 open source models (2025)
With the latest shift of US foreign policy, maybe we can start favouring European projects, like Mistral in France…
Open source technology in the age of AI (2025), by McKinsey
With more organizations deploying gen AI across business functions, a new survey finds that leaders are increasingly turning to open source AI solutions to build out their tech stacks.
Challenges and limits of an open source approach to Artificial Intelligence (2021)
A little bit old study by the European Parliament, but still interesting.
Tool Tip
A whole repository of “awesome AI apps” put together by Nina Duran.
In her own words:
💡 Discover practical and creative ways LLMs can be applied across different domains, from code repositories to email inboxes and more.
🔥 Explore apps that combine LLMs from OpenAI, Anthropic, Gemini, and open-source alternatives with AI Agents, Agent Teams, MCP & RAG.
🎓 Learn from well-documented projects and contribute to the growing open-source ecosystem of LLM-powered applications.
It's interesting how you framed this beyond just the technical aspect; I genuinley hadn't considered the agency argument so deeply. You really hit on why fostering a builder's mindset is so crucial for navigating the future of AI, and I couldn't agree more with your points on literacy and internal capability.